The impact of smoking and pre-existing illness on the relationship betweenbody-mass index and mortality

Purpose: To investigate the influence of smoking and pre-existing illness o n the relationship between body-mass index (BMI) and mortality from all cau ses, cardiovascular disease (CVD) and cancer. Methods: Study subjects were 168,828 men and 290,149 women, 45-64 years of age who participated in the C ancer Prevention Study I with 12 years of follow-up. We examined participan ts in four strata: (1) healthy never smokers, (2) healthy current smokers, (3) unhealthy never smokers and (4) unhealthy current smokers. Cox proporti onal hazard models with a piece-wise quadratic spline with 8 knots for BMI were used to plot the BMI-mortality relationship. Interactions of smoking a nd pre-existing illness with BMI were tested. Results: Three-way interactio ns between BMI, smoking and pre-existing illness were statistically signifi cant for all 3 mortality outcomes for both men and women. Healthy never-smo king participants tended to have the lowest hazard ratios at low BMI levels and the highest hazard ratios at high levels of BMI. Pre-existing illness tended to modify this effect more than smoking status. Conclusion: Pre-exis ting illness and smoking modify associations between BMI and mortality and should be carefully considered in an analysis of the BMI-mortality associat ion. (C) 2000 Elsevier Science Inc.